In an early stage of learning to write, diphthongs are complex structures that cause difficulties for many children. In addition to having a two-character representation, diphthongs have different phonetic forms that seem to interfere with children’s consolidation of the conventional spelling, depending on whether their dialect more or less resembles the graphic form of the structure. Based on the idea that, unlike the northern dialect, the southern dialect produces the diphthong /eI/ as [e], the goal of the present paper is to analyze the behavior of children from Alentejo and Trás-os-Montes (from Elvas (E), Vila Nova de Santo André (VNSA), Bragança (B) and Chaves (Ch)) from the 2nd year of schooling regarding this structure. It is also known that /eI/ centralization occurs in some regions beyond Lisbon, and its effect on children’s writing outside the capital is unknown. As such, written productions from the mentioned localities will be observed in order to compare the results with those from Lisbon (L) and Porto (P) of the same diphthong, reported by Rodrigues & Lourenço (2017). Our results concerning writing indicate that children from the South of the country (L, E + VNSA) have more difficulty in stabilizing the spelling of the diphthong <ei> than children from the North (P, B + Ch). The order of the hit rates is as follows: P >> B + Ch >> E + VNSA >> L, which confirms the existence of possible effect of dialect on the spelling performance of these children. The non-conventional forms (FNCs) varied according to dialectal region. In the two Alentejo localities, there was a preference for the erroneous forms <*e>/<*é> and <*ai>, which allows us to hypothesize that, in addition to the to the single vowel [e] form already described by dialectal studies, the oral form [ɐj] is also present there. In turn, in the localities of Bragança and Chaves, the FNC <*ai> was the most frequent, suggesting that the pronunciation in this region may be [ɐj] in addition to [ej]. Only the oral productions of the students who presented FNCs were analyzed, to try to understand whether the FNCs adopted derived from a direct transposition from their orality to writing. The observed orality data revealed that (i) there are children who establish a direct relation between oral form produced and FNC adopted, (ii) there are children whose oral and written production have no relation, thus, there is no transposition in writing of the oral form and, finally, (iii) there are children whose production in writing and/or orality varies depending on the lexical item. These results suggest that the analysis of children’s writing and speech may be a way to discover new clues for the development of other sociolinguistic work on Portuguese.